Complex Superstructures of Mo2P4O15 - Inorganic Chemistry (ACS

Complex Superstructures of Mo2P4O15. Sarah E. Lister, Ivana Radosavljević Evans and John S. O. Evans*. Department of Chemistry, Durham University, So...
2 downloads 0 Views 3MB Size
Inorg. Chem. 2009, 48, 9271–9281 9271 DOI: 10.1021/ic901090p

Complex Superstructures of Mo2P4O15 Sarah E. Lister, Ivana Radosavljevic Evans, and John S. O. Evans* Department of Chemistry, Durham University, South Road, Durham, DH1 3LE, United Kingdom Received June 5, 2009

We report structural studies on Mo2P4O15 over the temperature range 16-731 K, which show that it is considerably more complex than revealed by earlier work. Its low-temperature structure has lattice parameters a = 24.1134(6) A˚, b = 19.5324(5) A˚, c = 25.0854(6) A˚, β = 100.015(1)°, and V = 11635.0(5) A˚3 at 120 K, containing 441 unique atoms in space group Pn, a remarkably high number for a material with such a simple composition. Mo2P4O15 undergoes a structural phase transition at ∼520 K to a high-temperature phase in space group P1, with lattice parameters a = 17.947(3) A˚, b = 19.864(3) A˚, c = 21.899(3) A˚, R = 72.421(3)°, β = 78.174(4)°, γ = 68.315(4)°, and V = 6877.2(19) A˚3 at 573 K. The high-temperature structure, with 253 unique atoms, retains much of the low-temperature complexity.

Introduction Framework inorganic materials made up of corner-sharing metal-centered tetrahedra or octahedra show a wealth of fascinating properties. Zeolites, for example, are of use for ion exchange, gas sorption, catalysis, and many other applications.1 Nasicon (Na1þxZr2(PO4)3-x(SiO4)x) and structurally related phases show high ionic conductivity, and their chemical flexibility has been exploited for radioisotope capture and storage.2 Other framework oxides are of interest as intercalation hosts for rechargeable lithium batteries.3-5 We have a particular interest in the unusual thermal expansion properties and phase transitions of framework structures.6,7 One way of categorizing these materials is by the ratio of tetrahedral (T) to octahedral (M) cation sites they contain. An all-tetrahedral corner-linked oxide would have formula TO2 [TO4/2] and is represented by the many polymorphs of SiO2 and high silica zeolites, several of which show negative thermal expansion.8,9 An all-octahedral framework would be *To whom correspondence should be addressed. E-mail: john.evans@ durham.ac.uk. (1) Wright, P. A. Microporous Framework Solids; RSC: Cambridge, U. K., 2008. (2) Scheetz, B. E.; Agrawal, D. K.; Breval, E.; Roy, R. Waste Manage. 1994, 14, 498–505. (3) Padhi, A. K.; Nanjundaswamy, K. S.; Goodenough, J. B. J. Electrochem. Soc. 1997, 144, 1188–1194. (4) Tarascon, J. M.; Armand, M. Nature 2001, 414, 359–367. (5) Armand, M.; Tarascon, J. M. Nature 2008, 451, 652–657. (6) Evans, J. S. O. J. Chem. Soc., Dalton Trans. 1999, 19, 3317–3326. (7) Stinton, G. W.; Hampson, M. R.; Evans, J. S. O. Inorg. Chem. 2006, 45, 4352–4358. (8) Barrera, G. D.; Bruno, J. A. O.; Barron, T. H. K.; Allan, N. L. J. Phys.: Condens. Matter 2005, 17, R217–R252. (9) Colligan, M.; Forster, P. M.; Cheetham, A. K.; Lee, Y.; Vogt, T.; Hriljac, J. A. J. Am. Chem. Soc. 2004, 126, 12015–12022. (10) Dapiaggi, M.; Fitch, A. N. J. Appl. Crystallogr. 2009, 42, 253–258.

r 2009 American Chemical Society

MO3 [MO6/2], and NTE has again been reported in ReO3.10,11 Between these two extremes, we have the 2T:1M MT2O7 [(MO6/3)(TO4/2)2] materials such as ZrP2O7 and ZrV2O7, the large family of 3T:2M M2T3O12 [(MO6/2)2(TO4/2)3] materials, and 1T:1M materials such as NbOPO4 [(MO6/2)(TO4/2)].12-18 Members of all of these families show negative thermal expansion. These examples all contain fully cornerlinked polyhedra, and the requirement for T-O-T or M-O-M linkages within them varies systematically as the T/M ratio changes. For a stoichiometry of MT2O7, one has T2O7 units containing a T-O-T linkage, which in turn link to MO6 octahedra. At 3T:2M (where 3  4 corners of a tetrahedron match 2  6 corners of an octahedron), neither T-O-T or M-O-M links are required, and beyond this limit, M-O-M links are needed. In addition to the many polyhedral arrangements possible within each such scheme, structural possibilities are widened further if one considers frameworks in which nonlinked corners are allowed. If, for example, we introduce an additional oxygen into the structure of cubic MT2O7 and allow one oxygen of each TO4 to become one-coordinate, we obtain the structure of ZrW2O8, which can be described as corner-sharing ZrO6 octahedra (11) Chatterji, T.; Henry, P. F.; Mittal, R.; Chaplot, S. L. Phys. Rev. B 2008, 78, 134105. (12) Khosrovani, N.; Korthuis, V.; Sleight, A. W.; Vogt, T. Inorg. Chem. 1996, 35, 485. (13) Khosrovani, N.; Sleight, A. W.; Vogt, T. J. Solid State Chem. 1997, 132, 355–360. (14) Korthuis, V.; Khosrovani, N.; Sleight, A. W.; Roberts, N.; Dupree, R.; Warren, W. W. Chem. Mater. 1995, 7, 412–417. (15) Evans, J. S. O.; Mary, T. A.; Sleight, A. W. J. Solid State Chem. 1997, 133, 580. (16) Sleight, A. W.; Brixner, L. H. J. Solid State Chem. 1973, 7, 172. (17) Amos, T. G.; Sleight, A. W. J. Solid State Chem. 2001, 160, 230–238. (18) Amos, T. G.; Yokochi, A.; Sleight, A. W. J. Solid State Chem. 1998, 14, 303–307.

Published on Web 09/10/2009

pubs.acs.org/IC

9272 Inorganic Chemistry, Vol. 48, No. 19, 2009 and WO4 tetrahedra, with one oxygen per WO4 tetrahedron being formally one-coordinate. This family shows a wealth of fascinating properties related to its unusual topology.19-28 In this paper, we report structural studies on the Mo(V)containing material Mo2P4O15 (MoP2O7.5), which again has a stoichiometry related to cubic MP2O7 phases but now contains 0.5 extra oxygen atoms per formula unit. Its composition is thus midway between MT2O7 and MT2O8 materials. The Mo(V) valence state has been confirmed by electron paramagnetic resonance.29 There have been a number of structural studies on this phase, and the basic polyhedral connectivity is well established.30-33 The structure contains MoO6 octahedra and PO4 tetrahedra with tetrahedra cornerlinked to form P4O13 groups (Figure 1). These contain three “bridging” P-O-P oxygens, and the remaining 10 are shared with Mo sites. The MoO6 octahedra are completed by a one-coordinate or “terminal” oxygen, Figure 1, such that the formula is perhaps better expressed as (MoO)2P4O13. In contrast to ZrW2O8, the one-coordinate oxygen is located on the octahedral rather than the tetrahedral atom. While this basic connectivity is well established, there has been some debate about the precise details of the structure. Unit-cell parameters that have been reported to date are summarized in Table 1. The various cells are clearly closely related, with each volume a simple multiple of the smallest cell (hereafter, the P21/c subcell) used by Costentin et al. In the Costentin et al. publication, analysis of single-crystal data gave R/Rw values of 3.6/3.7% using a simple model containing 11 atoms in the asymmetric unit, though it was necessary to disorder the central P-O-P bridging oxygen atom of P4O13 groups over two sites close to a center of inversion to achieve this fit, and the authors acknowledged that anisotropic displacement parameters (adp’s) were unusually large. We have previously reported the structure of this material at 120 K.33 In this paper, we report structural studies on Mo2P4O15 over a temperature range of 16-731 K which reveal that it is considerably more complex than suggested by earlier work over this entire range. Its low-temperature (19) Mary, T. A.; Evans, J. S. O.; Vogt, T.; Sleight, A. W. Science 1996, 272, 90–92. (20) Ernst, G.; Broholm, C.; Kowach, G. R.; Ramirez, A. P. Nature 1998, 396, 147. (21) Ramirez, A. P.; Kowach, G. R. Phys. Rev. Lett. 1998, 80, 4903. (22) Perottoni, C.; da Jornada, J. A. H. Science 1998, 280, 886–889. (23) Hancock, J. N.; Turpen, C.; Schlesinger, Z.; Kowach, G. R.; Ramirez, A. P. Phys. Rev. Lett. 2004, 93, 225501. (24) Drymiotis, F. R.; Ledbetter, H.; Betts, J. B.; Kimura, T.; Lashley, J. C.; Migliori, A.; Ramirez, A. P.; Kowach, G. R.; Van Duijn, J. Phys. Rev. Lett. 2004, 93, 025502. (25) Kennedy, C. A.; White, M. A. Solid State Commun. 2005, 134, 271– 276. (26) Pantea, C.; Migliori, A.; Littlewood, P. B.; Zhao, Y.; Ledbetter, H.; Lashley, J. C.; Kimura, T.; Van Duijn, J.; Kowach, G. R. Phys. Rev. B 2006, 73, 214118. (27) Keen, D. A.; Goodwin, A. L.; Tucker, M. G.; Dove, M. T.; Evans, J. S. O.; Crichton, W. A.; Brunelli, M. Phys. Rev. Lett. 2007, 98, 225501. (28) Dubbeldam, D.; Walton, K. S.; Ellis, D. E.; Snurr, R. Q. Angew. Chem., Int. Ed. 2007, 46, 4496–4499. (29) Lezama, L.; Rojo, J. M.; Pizarro, J. L.; Arriortua, M. I.; Rojo, T. Solid State Ionics 1993, 63-65, 657. (30) Schulz, I. Z. Anorg. Allg. Chem. 1955, 281, 99–112. (31) Minacheva, L. K.; Antsyshkina, A. S.; Lavrov, A. V.; Sakharova, V. G.; Nilolaev, V. P.; Porai-Koshits, M. A. Rus. J. Inorg. Chem. 1979, 24, 51–53. (32) Costentin, G.; Leclaire, A.; Borel, M.-M.; Grandin, A.; Raveau, B. Z. Kristallogr. 1992, 201, 53–58. (33) Lister, S. E.; Evans, I. R.; Howard, J. A. K.; Coelho, A. A.; Evans, J. S. O. Chem. Commun. 2004, 22, 2540–2541.

Lister et al.

Figure 1. Portions of the structure of Mo2P4O15 (a) emphasizing the local environment around each MoO6 octahedron and (b) a P4O13 unit made from corner-sharing PO4 groups.

structure contains 441 atoms in the asymmetric unit, a remarkably high number for a material with such a simple composition. Mo2P4O15 undergoes a structural phase transition at ∼520 K, but its high-temperature structure retains much of the low-temperature complexity. Experimental Section Mo2P4O15 was prepared following the method of Lezama et al.29 A total of 1.4395 g of MoO3 (10 mmol, Alfa, 99.95%) was intimately mixed with (NH4)2HPO4 (30 mmol, Aldrich, 98þ %) and warmed at 5 K min-1 to 523 K, held at this temperature for 2 h, heated at 2 K min-1 to 873 K, held at this temperature for 6 h, and then cooled by turning off the furnace. The resulting green solid was washed with demineralized water and dried to produce 2.1981 g (79% yield) of Mo2P4O15 (PDF 76-2079) containing a number of crystals of suitable size for single-crystal experiments. High-temperature powder diffraction experiments were performed using Cu KR1 radiation on a Bruker d8 diffractometer equipped with a Ge(111) incident beam monochromator, a Vantec psd, and an Anton Paar HTK1200 heating stage using a sample mounted on an amorphous silica disk. To check temperature calibration, the sample was mixed with an Al2O3 internal standard.34 A total of 309 30-min scans from 10 to 90° 2θ with a step size of 0.017° 2θ were recorded in 5 K steps from 308 to 731 K in two cycles of heating and cooling. Excellent reproducibility was obtained between cycles. Low-temperature experiments were performed using KR1/ KR2 radiation on a Bruker d8 diffractometer equipped with a Lynxeye psd and an Oxford Cryosystems pHeniX cryostat. The sample was mixed with a Si internal standard and sprinkled on an Al plate. Peak positions of both Si and Al were used to verify temperature calibration.34 Data sets were collected from 10 to 90° 2θ in step sizes of 0.017° 2θ over 20 min time slices as the sample was cooled at 17 K hr-1 from 300 to 16.4 K then warmed from 16.4 to 300 K. Powder diffraction data were analyzed using a combination of Rietveld and Pawley refinements to extract the temperature dependence of cell parameters. For high-temperature refinements, a total of 67 parameters were refined for each data set (4 cell parameters, 24 terms of a spherical harmonic preferred orientation correction, 3 adp’s, 4 peak shape parameters, and a scale factor for Mo2P4O15; 2 cell parameters, 2 atomic coordinates, 2 adp’s, 4 peak shape parameters, and a scale factor for the Al2O3 standard; 18 background parameters, a sample height correction, and 1 parameter to describe axial divergence). For low-temperature data, 74 parameters were refined (4 cell parameters, 24 terms of a spherical harmonic preferred orientation correction, 3 adp’s, 6 peak shape parameters, and a scale factor for Mo2P4O15; 1 cell parameter, 1 adp, 6 peak shape parameters, (34) Stinton, G. W.; Evans, J. S. O. J. Appl. Crystallogr. 2008, 40, 87–95.

Article

Inorganic Chemistry, Vol. 48, No. 19, 2009

9273

Table 1. Unit Cell Parameters and Space Groups Previously Reported for Mo2P4O15 at Room Temperature and belowa reference Schulz30 Minacheva et al.31 Costentin et al.32 this work33 a

space group Pc P21/c Pn

a/A˚

b/A˚

c/A˚

β/°

V/A˚3

V/Vsub

≈ 16.7 8.288(2) 8.3068(8) 24.133(2)

≈ 19.3 19.529(5) 6.5262(6) 19.579(2)

≈ 10.8 10.690(3) 10.7181(11) 25.109(2)

≈ 107.5 106.7(3) 106.7050(78) 99.962(3)

≈ 3320 1657.18 556(1) 11685(1)

6 3 1 21

Vsub refers to the volume of the cell reported by Costentin.

Figure 2. X-ray and neutron diffraction data showing the phase transition in Mo2P4O15. Data presented as a pseudo-film plot using Powder 3D software.39 The light band in the neutron data is caused by a shorter effective collection time at this temperature.

and a scale factor for Si; 1 cell parameter, 5 peak intensities, a height offset parameter, and 6 peak shape parameters for Al; 12 background parameters, a sample height correction, and 1 parameter to describe axial divergence). Mo2P4O15 and Si were modeled using the Rietveld method, whereas the textured Al sample holder was fitted using the Pawley method. These protocols gave excellent fits to data over the whole temperature range, and checks were made to ensure that other parameters did not correlate significantly with the key unit-cell parameters we were trying to extract. All refinements were performed using the Topas Academic software suite controlled by local routines.35 Neutron powder diffraction experiments were performed using the HRPD diffractometer at the ISIS facility of the Rutherford Appleton Laboratory on a 6.03 g sample mounted in a 15 mm V can filled to a depth of 30 mm. Eight batches of material were combined to make this neutron sample. Data were collected over a time-of-flight range of 10-210 ms from 303 to 653 K in 10 K steps collecting 4 μAh (15 min) at each temperature. Due to a problem with detector electronics, data between 493 and 543 K were not fully recorded. When the sample reached 543 K, it was therefore recooled to 503 K, and data collection resumed. The effective data collection time at 493 K was therefore considerably shorter, and a small discontinuity is observed in the data which is not an intrinsic property of the sample. The 120 and 293 K single crystal experiments were performed using a Bruker AXS Smart 6000 diffractometer with Mo KR1 radiation. Temperature control was achieved using an Oxford Cryosystems cryostream. A total of 4140 frames were collected in ω steps of 0.3° using a data collection time of 30 s and integrated using the SAINT software package.36 Data were collected at generator settings of 35 kV and 50 mA (35) Coelho, A. A. TOPAS Academic: General Profile and Structure Analysis Software for Powder Diffraction Data; Bruker AXS: Karlsruhe, Germany, 2004. (36) SAINT, V6.02; Bruker AXS: Madison, WI, 2005.

to avoid problems with λ/2 contamination and corrected for absorption using the SADABS program.37 Structure solution using simulated annealing methodologies was performed using the Topas Academic software suite35 and final refinement using Crystals.38 A second 293 K and 573 and 653 K data collections used an Enraf Nonius FR559 Crystal Heater. The heater attachment was mounted vertically on the diffractometer (χ = 0). A total of 600 frames were collected in ω steps of 0.3° using a data collection time of 10 s. A total of 20 209 reflections with I > 3σ(I) could be collected at 293 K using this configuration. No absorption correction was applied for data collected in this mode due to low redundancy. We note that successful refinement of the 293 K data led to verification of the use of this data collection strategy at high temperatures. Fewer reflections recorded at high temperatures (573 K, 7680 reflections; 653 K, 4834 reflections) led to higher standard uncertainties on refined values. A sample of Mo2P4O15 was examined using DSC, heating and cooling over the range 298-573 at 10 K/min, using a Perkin-Elmer Pyris 1 Differential Scanning Calorimeter.

Results Thermal Expansion of Mo2P4O15. Variable-temperature powder X-ray diffraction data were collected on a phase-pure sample of Mo2P4O15 from 16.4 to 731 K. Data collected above room temperature are included in Figure 2. Unit-cell parameters were extracted by Rietveld refinement and sample temperatures calibrated using internal standards. The supercell model was used to fit all experimental data to allow volume evolution to be followed over the whole temperature range of cells; the :: :: (37) Sheldrick, G. M. SADABS; University of Gottingen: Gottingen, Germany, 1996. (38) Betteridge, P. W.; Carruthers, J. R.; Cooper, R. I.; Prout, K.; Watkin, D. J. J. Appl. Crystallogr. 2003, 36, 1487. (39) Hinrichsen, B.; Dinnebier, R. E.; Jansen, M. Powder 3D, v1.1; Stuttgart, Germany, 2004.

9274 Inorganic Chemistry, Vol. 48, No. 19, 2009

Lister et al.

Figure 3. Temperature evolution of unit-cell parameters a, b, c, and β and volume and volume thermal expansion coefficients for the Pn supercell model. Legend (a-e): (red/blue 2) low-temperature heating/cooling cycles, (red/blue 9) furnace heating/cooling cycle 1, (red/blue b) furnace heating/cooling cycle 2, (black 2) single-crystal data collections; (f) (blue/gray () thermal expansion data.

cell volume plotted above ∼ 520 K is therefore 1.75 times the true volume. The temperature dependence of unit-cell parameters a, b, c, β, and V are shown in Figure 3. Cell parameters derived from single-crystal experiments are included in Figure 3a-e for comparison and show good agreement with powder values. The temperature dependence of the volume thermal expansion coefficient (Figure 3f) was determined by fitting a Bezier function to the cell volume data followed by numerical differentiation. This procedure means the first few data points are relatively unreliable, but they are included in the figure in gray scale for completeness. Figure 3f shows that Mo2P4O15 displays negative thermal expansion below ∼80 K. This is presumably due to population of negative-Gruneisen-parameter phonon modes at low temperatures involving local transverse vibrations of two-coordinate oxygen, as observed in other NTE framework structures. Above ∼80 K, smooth positive expansion of all cell parameters is observed up to a sharp discontinuity in values at ∼520 K, consistent with a first-order phase transition involving a volume change of ΔV ∼ þ1%. An approximately 10 K hysteresis in phase

transition temperature is observed on cooling. Percentage changes in a, b, and c over the whole temperature range of þ0.58, þ1.68, and þ1.0% are observed. Similar phase transition temperatures of 540 and 523 K on warming and cooling were obtained by differential scanning calorimetry; these data are included in the Supporting Information, Figure S1. Powder diffraction data recorded above and below the phase transition temperature showed relatively minor changes in the X-ray data but more marked changes in the intensities of the neutron data (Figure 2), suggesting that the changes occurring are displacive in nature and mainly affect the oxygen atoms. Despite the simplification of neutron diffraction data at high temperatures (in that fewer superstructure reflections are observed), reflections are still present which cannot be indexed using the P21/c subcell. Low-Temperature Structures of Mo2P4O15. We have briefly reported the low-temperature structure of Mo2P4O15 (hereafter referred to as R-Mo2P4O15 or the Pn structure) in an earlier communication.33 Diffraction images recorded at 120 K required a unit cell considerably larger than the 556 A˚3 monoclinic cell reported by Costentin et al.

Article

Figure 4. Comparison of the Costentin P21/c model with Pn and P1 structures of Mo2P4O15. All structures viewed down the b axis. Pn and P1 structures have b axes 3 times that of the Costentin model.

but could instead be indexed using a monoclinic cell with a = 24.133 A˚, b = 19.579 A˚, c = 25.109 A˚, and β = 99.962°. The unit-cell volume (11685 A˚3) is 21 times that of the smaller cell, and the two cells can be related by the transformation (3 0 1, 0 3 0, -1 0 2), as shown in Figure 4. An automated analysis of systematic absences suggested space group P21/n, though careful inspection showed that reflections forbidden by the 21 axis were present though weak. The strongest 0 k 0 reflection with k 6¼ 2n was 0 7 0 with I = 1.54(3), which ranked 13352th of the observed reflections ordered by intensity, and was 0.02% of the strongest reflection [I = 9300(120)]. The cell volume and symmetry imply 441 atoms in the asymmetric unit. Due to the significant pseudosymmetry present, the superstructure was initially solved using a simulated annealing methodology in the Topas software suite. Coordinates in the large cell were generated from those reported in the Costentin paper, and all atomic coordinates and a single overall temperature factor were allowed to refine to convergence from their ideal starting values using a limited data range (0.4 A˚. On convergence, atomic coordinates were reset to values randomly displaced from “ideal” positions by up to 0.1 A˚ and the structure re-refined. The best solution after several rounds of refinement/perturbation gave good agreement with the diffraction data. Isotropic refinement using the whole data set led to a final R factor of 3.49% for 43 738 reflections with I > 3σ(I). Scatter plots of Iobs versus Icalcd showed no systematic deviations for weak reflections. We note that recently developed charge flipping methods of structure solution are particularly tolerant of pseudosymmetry.41,42 In fact, a “default” application of charge flipping as implemented in the Topas software suite43 will typically correctly locate all (or the vast majority of) atomic positions within a few seconds, producing a trial structure which can be refined readily. (40) Costentin, G.; Leclaire, A.; Borel, M. M.; Grandin, A.; Raveau, B. Rev. Inorg. Chem. 1993, :: 77–101. :: 13, (41) Oszlanyi, G.; Suto, A. Acta Crystallogr., Sect. A 2004, 60, 134–141. (42) Oszlanyi, G.; Suto, A.; Czugler, M.; Parkanyi, L. J. Am. Chem. Soc. 2006, 128, 8392–8393. (43) Coelho, A. A. Technical Reference, Topas Manual, v4.1; Coelho Software: Brisbane, Australia, 2007.

Inorganic Chemistry, Vol. 48, No. 19, 2009

9275

This procedure was therefore adopted for later hightemperature data sets. Histograms of bond distances for the 42 crystallographically unique MoO6 octahedra, 84 unique PO4 tetrahedra, and P-O-P bridging bond angles in the structure are shown in Figure 5 alongside equivalent plots for the high-temperature structure. These will be discussed below. Data were also collected at 293 K in two separate experiments: one repeating the 120 K data collection strategy, and one using the setup employed for collections above room temperature. Correlation of bond distances and angles in the 120 and 293 K structures showed no significant changes over this temperature range, with similar spreads of bond distances and angles. The 293 K structures using the two strategies showed minimal differences. High-Temperature Structures of Mo2P4O15. High-temperature data were recorded using an Enraf Nonius heater attachment mounted on a Bruker AXS Smart 6000 at set temperatures of 573 and 653 K. Temperature calibration was checked both using an external thermocouple and by comparing unit-cell parameters to those obtained by powder diffraction methods; temperatures are estimated to be correct to (10 K. While considerably fewer reflections were observed above the 520 K phase transition than below, significantly more reflections than those predicted by the 556 A˚3 subcell were observed. This is consistent with neutron powder diffraction data, which also showed extra reflections. Diffraction data at both 573 and 653 K could be indexed with a triclinic cell with a volume ∼12 times that of the subcell (or 4/7 that of the true low-temperature cell), with 573 K values of a = 17.947 A˚, b = 19.864 A˚, c = 21.899 A˚, R = 72.421°, β = 78.174°, γ = 68.315°, and V = 6877 A˚3. This can be related to the P21/c subcell by the transformation matrix (-2 1 0, 0 3 0, -1 1 -2). The cell size and symmetry again imply a complex structure which would require 24 Mo, 48 P, and 181 O atoms (assuming, see below, that two oxygens lie on 1 sites) or 253 atoms in total in the asymmetric unit. The structure of the high-temperature form (β-Mo2P4O15) was solved using charge-flipping methods using an applied symmetry of P1 and applying the tangent formula for phase refinement as described by Coelho.44 This procedure successfully located all Mo and P sites and the majority of the 181 oxygens. The remainder were located using Fourier methods. A model with the same connectivity as at low temperatures resulted, which could be refined to R/Rw = 5.64/10.02% for 7680 observations with I>2σ(I). The parameter/observation ratio (1011:7680) restricts us to using isotropic displacement parameters. The structure was essentially unchanged at 653 K. Details of this refinement are given in Table 2, and the CIF is included in the Supporting Information. Examination of histograms of bond distances and angles revealed a significant increase in spread relative to 120 or 293 K refinements, though mean values (Table 3) were within a standard uncertainty of those at lower temperatures. The lack of significant thermal expansion of bonds is a common observation in framework oxides and reflects the artificial reduction in bond lengths due to correlated (44) Coelho, A. A. Acta Crystallogr., Sect. A 2007, 63, 400–406.

9276 Inorganic Chemistry, Vol. 48, No. 19, 2009

Lister et al.

Figure 5. Histograms showing distributions of Mo-O (a,b) and P-O (c,d) bond lengths, P-O-P bond angles (e,f), and isotropic adp’s (g,h) in 120 K (left column) and 573 K (right column) temperature forms of Mo2P4O15. The y axes scales for histograms a-f reflect the 7:4 volume ratio between the two structures.

local motion compensating real expansion. The increased spread of distances is unsurprising given the significant reduction in the number of observed reflections and the higher parameter/observation ratio. For final refinements, bond distances within polyhedra were therefore restrained to values obtained at 120 K, with different values for “long”, “medium”, and “short” bonds within MoO6

octahedra and for P-O(-Mo) and P-O(-P) P-O bonds. Restraints were weighted according to the standard deviations of the 120 K distributions. Refinement led to final R factors of R/Rw = 6.15/10.73%, only marginally higher than those for the unrestrained model. Histograms of bond distances and angles are included in Figure 5 for comparison with the low-temperature structure.

Article

Inorganic Chemistry, Vol. 48, No. 19, 2009

9277

Table 2. Crystallographic Data for the Superstructure Phases of Mo2P4O15 Discussed in This Paper 120 K

293 K

chemical formula Mo2P4O15 Mo2P4O15 Mr 555.77 555.77 cryst syst monoclinic monoclinic space group (No.) P1n1 (No. 7) P1n1 (No. 7) temp/K 120 293 a/A˚ 24.1134 (6) 24.133 (2) b/A˚ 19.5324 (5) 19.5793 (18) c/A˚ 25.0854 (6) 25.109 (2) R/° 90 90 β/° 100.0150 (10) 99.962 (3) γ/° 90 90 11635.0 (5) 11685.4 (19) V/A˚3 Z 21 21 -3 3.331 3.317 Dcalcd/g cm λ/A˚ 0.71073 0.71073 2.93 2.92 μ/mm-1 no. of measured, independent, and 93645, 48060, 43783 216594, 48901, 9315 obsd reflns criterion for obsd reflns I > 3.0σ(I ) I > 3.0σ(I ) 0.035, 0.060, 0.86 0.033, 0.083, 0.79 R(F2),aRw(F2),bS no. of params 1766 1766 observations: parameters ratio 24.8 5.3 P P P P a R = ||Fo| - |Fc||/ |Fo|. b Rw = [ ||Fo2| - |Fc2||2/ w|Fc2|2]1/2.

293 K (hothead)

573 K (restrained)

653 K (restrained)

Mo2P4O15 555.77 monoclinic P1n1 (No. 7) 293 24.1291 (10) 19.5817 (8) 25.1108 (10) 90 99.9780 (10) 90 11685.1 (8) 21 3.317 0.71073 2.92 44022, 30516, 20209

Mo2P4O15 555.77 triclinic P1 (No. 2) 573 17.947 (3) 19.864 (3) 21.899 (3) 72.421 (3) 78.174 (4) 68.315 (4) 6877.2 (19) 12 3.220 0.71073 2.84 25767, 25767, 7680

Mo2P4O15 555.77 triclinic P1 (No. 2) 653 18.001(17) 19.98(3) 22.00(3) 72.19(5) 78.24(6) 68.42(4) 6967(16) 12 3.179 0.71073 2.80 26002, 26002, 4834

I > 3.0σ(I ) 0.048, 0.115, 0.99 1766 11.4

I > 2.0σ(I ) 0.061, 0.107, 1.12 1011 7.6

I > 3.0σ(I ) 0.050, 0.128, 1.07 1011 4.78

Table 3. Average Mo-O and P-O Bond Lengths in Refinements Discusseda

Mo-O (short) Mo-O (medium) Mo-O (long) P-O(-Mo) P-O(-P)

120 K

293 K

293 K (hothead)

573 K free

573 K restrained

1.653(5) 2.025(25) 2.167(22) 1.500(13) 1.583(14)

1.652(26) 2.025(35) 2.172(35) 1.497(39) 1.583(36)

1.647(11) 2.022(26) 2.168(26) 1.497(16) 1.580(18)

1.654(100) 2.041(94) 2.171(89) 1.499(83) 1.582(91)

1.652(9) 2.035(36) 2.165(24) 1.502(14) 1.580(16)

a

293 K (hothead) refers to a 293 K refinement using the same data collection strategy used at 573 and 653 K. Numbers in parentheses are standard deviations of the set of individual distances in each category.

The unit cell size and space-group symmetry of β-Mo2P4O15 require that two oxygen atoms lie on a center of inversion;those labeled O206 and O212. This implies that two P-O-P angles out of 36 in the structure are required to be 180° or that two of the 12 independent P4O13 groups have an unfavorable local geometry. Figure 6 shows isotropic displacement parameters for central P2O7 units within these two P4O13 groups and Fourier maps calculated in an O-P-O-P-O plane; these suggest that the behavior of these two oxygens differ. O212 appears to be best described as being disordered over two sites close to the inversion center, whereas O206 appears adequately described as lying on the inversion center with thermal displacement around this position. The O212 site was therefore split in final refinements, and the local P-O-P angle derived was ≈ 137°. An angle close to 180° for O206 is not unreasonable given the P-O-P angle distribution observed at low temperatures (Figures 5 and 7). Discussion Examination of the histograms of bond distances at 120 K in Figure 5 suggests a high-quality refinement. Given the connectivity of the polyhedra, one would expect that the 42 crystallographically unique MoO6 octahedra would contain 42 “short” bonds to a terminal oxygen, 168 “medium” bonds to equatorial oxygens, and 42 “long” bonds to the oxygen trans to the terminal one. The histogram shows precisely this

distribution. Similarly for the 84 PO4 tetrahedra making up P4O13 groups, one would expect 210 P-O(-Mo) bonds around 1.50 A˚ and 126 longer P-O(-P) bonds around 1.58 A˚;exactly this distribution is again observed. Average bond lengths and their standard deviations are included in Table 3. Bond valence sums for all Mo and P sites deviate by 100 crystallographically unique atoms and only ∼100 more than 250. For the majority of these materials, it is relatively straightforward to identify the origin of their complexity. (48) ICSD, v 2008/2; Fachinformationszentrum Karlsruhe and the NIST: Gaithersburg, MD, 2005. (49) Weber, T.; Dschemuchadse, J.; Kobas, M.; Conrad, M.; Harbrecht, B.; Steurer, W. Acta Crystallogr., Sect. B 2009, 65, 308–317.

We discuss this in outline below for all materials with >100 unique atoms and in more detail for the handful of “nonmolecular” structures in this category. Details of the structures discussed are given in the Supporting Information, Table S3. For commensurate structures, the most obvious origin of structural complexity is chemical complexity: if one has a complex composition or a complex pattern of strong bonds or structure-directing weak interactions in a material, the structure will necessarily be complex. Protein structures fall into this category, as do a large number of recently reported polyoxometallates grown from solution. In the 2008/2 ICSD release, the six most complex (on our limited definition) structures :: come from Muller and co-workers, with the most complex being Na15[MoVI126MoV28O462H14(H2O)70]0.5[MoVI124MoV28O457H14(H2O)68]0.5 3 ∼400H2O with 937 atoms in the asymmetric unit.50-55 Looking at the 100 most complex structures, ∼85% are polyoxometallates, and a further ∼5% have complex compositions containing polyatomic cations or anions and are prepared under nonthermodynamic conditions. The most complex material which we would view as nonmolecular is (for this discussion we categorize the polyoxometallates grown from solution-based species as molecular in nature) Ba45Cu28Al17F197 with 574 unique atoms.56 A second source of complexity can be associated with materials prepared at high temperatures and which undergo (50) Mueller, A.; Das, S. K.; Fedin, V. P.; Krickemeyer, E.; Beugholt, C.; Boegge, H.; Schmidtmann, M.; Hauptfleisch, B. Z. Anorg. Allg. Chem. 1999, 625, 1187–1192. (51) Cronin, L.; Beugholt, C.; Krickemeyer, E.; Schmidtmann, M.; Boegge, H.; Koegerler, P.; Kim, T.; Luong, K.; Mueller, A. Angew. Chem., Int. Ed. Engl. 2002, 41, 2805–2808. (52) Mueller, A.; Krickemeyer, E.; Boegge, H.; Schmidtmann, M.; Peters, F.; Menke, C.; Meyer, J. Angew. Chem., Int. Ed. Engl. 1997, 36, 484–486. (53) Mueller, A.; Das, S. K.; Fedin, V. P.; Krickemeyer, E.; Beugholt, C.; Boegge, H.; Schmidtmann, M.; Hauptfleisch, B. Z. Anorg. Allg. Chem. 1999, 625, 1187–1192. (54) Mueller, A.; Beugholt, C.; Boegge, H.; Schmidtmann, M. Inorg. Chem. 2000, 39, 3112–3113. (55) Mueller, A.; Shah, S. Q. N.; Boegge, H.; Schmidtmann, M. Nature 1999, 397, 48–50. (56) Dupont, N.; Gredin, P.; Samouel, M.; de Kozak, A. Z. Anorg. Allg. Chem. 2002, 628, 191–197.

9280 Inorganic Chemistry, Vol. 48, No. 19, 2009

Lister et al.

Figure 9. Histogram showing the distribution of entries in the ICSD in 25-atom bins according to the number of atoms in the asymmetric unit. The main plot is presented on a logarithmic scale and the insert on a linear scale.

site ordering on cooling. One such example is R-La2Mo2O9 (312 atoms), which, apart from the title phase, is the most complex oxide in the ICSD.57 At high temperatures (>580 °C), this material shows high oxide ion conductivity,58 and its average structure can be described using a cubic cell with just five atoms in the asymmetric unit, though two of these are disordered. When cooling site-ordering of oxide ions occurs, Mo and La adopt a range of local coordination environments, and the room-temperature structure becomes monoclinic with 312 unique atoms. Atom and charge ordering is also the origin of complexity in the 376 atom structure proposed for the perovskite (Nd7/12Li1/4)TiO3 by Guiton et al.59 Here, one must accommodate both Nd/Li A site ordering and phase separation to Nd- and Li-rich regions on a nanometre length scale, which requires a 14ap  28ap  2ap cell (where ap is the cell parameter of a simple perovskite). Since only transmission electron microscopy and powder data are available on the material, no detailed refinement of the model has been attempted. Similarly, Sr4Ga2O7 (104 atoms) can be viewed as a 3  4  4 perovskite superstructure [Sr(Sr1/3Ga2/3)(O7/902/9)3] in which ordering the 22% oxygen vacancies and Ga sites gives rise to a mixture of GaO4 isolated tetrahedra and Ga3O10 corner-sharing trimers and requires 104 unique atoms.60 Ordering caused by stereochemically active lone pairs can also lead to complexity, as has been reported for materials such as Bi-containing distorted pyrochlores Bi2Sn2O7 and Bi2Hf2O7 (both 176 atoms).61,62 Nowogrocki et al. have reported the structure of Cs3B7O12 (219 atoms), which has a layered structure containing cornersharing BO3 triangles and BO4 tetrahedra with charge-balancing (57) Evans, I. R.; Howard, J. A. K.; Evans, J. S. O. Chem. Mater. 2005, 17, 4074–4077. (58) Lacorre, P.; Goutenoire, F.; Bohnke, O.; Retoux, R.; Laligant, Y. Nature 2000, 404, 856–858. (59) Guiton, B. S.; Hui, Wu; Davies, P. K. Chem. Mater. 2008, 20, 2860– 2862. (60) Kahlenberg, V.; Lazic, B.; Krivovichev, S. V. J. Solid State Chem. 2005, 178, 1429–1439. (61) Evans, I. R.; Howard, J. A. K.; Evans, J. S. O. J. Mater. Chem. 2003, 13, 2098–2103. (62) Henderson, S. J.; Shebanova, O.; Hector, A. L.; McMillan, P. F.; Weller, M. T. Chem. Mater. 2007, 19, 1712–1722.

Cs ions both within and between the slabs.63 Here, the complexity appears related to achieving stable local coordination environments for all tetrahedral BO4 and trigonalplanar BO3 units in the polyanionic network at this particular composition. Perhaps the most common examples of structural complexity in inorganic materials with simple compositions occur in frameworks based on corner-sharing polyhedra (such as the title compound), which undergo displacive phase transitions on cooling. Common examples where phase transitions lead to a moderate rise in complexity include the quartz and cristobalite polymorphs of SiO2, perovskites, and WO3. More complex structures occur for monoclinic and triclinic forms of AlPO4 with tridymite-related structures; by analogy with pure silica polymorphs, these materials are believed to contain 144 and 240 atoms in their asymmetric units at room temperature, though they have only been examined using heavily restrained Rietveld refinements.64 The most complex frameworks studied by single-crystal methods are ZrP2O7 (136 atoms, third most complex oxide)47 and Ru(PO3)3 (103 atoms),65 where the origin of complexity is again due to local bonding requirements of P-O-P linkages. There is also good evidence, for example, from powder diffraction and 31P NMR studies, that related materials such as SnP2O7 (540 atoms) are more complex still.66 Finally, Kahlenberg et al. have shown how the complex structure of γ-Na2Si2O5 (108 atoms) arises from both displacive phase transitions and site ordering.67 This material has an unusual structure based on Q3 SiO4 units that only share three of their four corners with other tetrahedra, such that it is a rare example of an interrupted 3D silicate (63) Nowogrocki, G.; Penin, N.; Touboul, M. Solid State Sci. 2003, 5, 795–803. (64) Graetsch, H. Acta Crystallogr., Sect. C 2000, 56, 401–403. (65) Imoto, H.; Fukuoka, H.; Tsunesawa, S.; Horiuchi, H.; Amemiya, T.; Koga, N. Inorg. Chem. 1997, 36, 4172–4181. (66) Fayon, F.; King, I. J.; Harris, R. K.; Gover, R. K. B.; Evans, J. S. O.; Massiot, D. Chem. Mater. 2003, 15, 2234–2239. (67) Kahlenberg, V.; Rakic, S.; Weidenthaler, C. Z. Kristallogr. 2003, 218, 421–431.

Article

framework and can in fact be viewed as a defect cristobalite network with 20% of the tetrahedral centers removed in an ordered fashion. Above ∼560 °C, the material has a relatively simple tetragonal structure which requires, at least on average, one 180° Si-O-Si angle. On cooling, a superstructure with 12 times the volume containing 108 unique atoms forms. The complexity of Mo2P4O15 is clearly related to these other framework materials. It is, however, unusual both in the large number of unique atoms at room temperature and since the high-temperature β form retains much of this complexity, requiring 253 atoms in the asymmetric unit and a 6877 A˚3 cell. In fact, the number of free structural parameters per unit volume is barely changed at the R-to-β transition. In the β structure, if both O206 and O212 lie on 1, the number of parameters per subcell unit volume would be 62.75; if O212 is off 1, then that number is 63. In Pn, the equivalent number is 63. This contrasts with many frameworks where high-temperature structures are considerably more symmetric than low, though this is not a thermodynamic requirement (for example, systems with re-entrant phase transitions such as Rochelle’s salt,68,69 Ag7P3S11,70 (68) Beevers, C. A.; Hughes, W. Proc. R. Soc. London. Ser. A 1941, 177, 251–259. (69) Suzuki, E.; Shiozaki, Y. Phys. Rev. B 1996, 53, 5217–5221. :: (70) Brinkmann, C.; Eckert, H.; Wilmer, D.; Vogel, M.; auf der Gunne, J. S.; Hoffbauer, W.; Rau, F.; Pfitzner, A. Solid State Sci. 2004, 6, 1077–1088.

Inorganic Chemistry, Vol. 48, No. 19, 2009

9281

and malononitrile71 involve a high-to-low symmetry transition on warming). In fact, under ambient conditions, Mo2P4O15 decomposes before it transforms to the expected simple P21/c subcell structure. The effective energy barriers for P-O-P inversion within the framework are therefore significantly higher than predicted by simple gas-phase calculations. Acknowledgment. We thank the EPSRC for funding under EP/C538927/1, STFC for access to ISIS central facilities, and the Royal Society for funding which allowed purchase of the single-crystal high-temperature attachment. SEL thanks Durham University for a Doctoral Training Fellowship. We thank Richard Ibberson and Kevin Knight for assistance with the neutron diffraction experiments. We also thank Bob McMeeking for providing information extracted from the ICSD allowing structures to be ranked according to the number of atoms in their asymmetric unit. Supporting Information Available: X-ray crystallographic files in CIF format for the Mo2P4O15 structures at 120, 293, 293 (hothead), and 573 K (hothead), 573 K (restrained) and 653 K (restrained). Tables of low-temperature Mo/P bond valence sum calculations. Table of complex nonmolecular materials in ICSD discussed in the text. This material is available free of charge via the Internet at http://pubs.acs.org. (71) Dove, M. T.; Rae, A. I. M. Faraday Discuss. 1980, 69, 98–106.